Real-time approach for the effect of arm position using Fitts’ Law

Abstract

Electromyography is a method or technique that is utilized to evaluate or record the electrical activity of muscles. These recorded signals can reveal some notable observations that can be used to enhance myo electric-based prosthetic devices. The prosthetic devices that are commonly available have changed the lives of people significantly with upper limb amputation. However, currently available devices are calibrated in a single position, which gives rise to problems that can affect these devices accuracy and efficiency. The “limb position effect” is a prominent problem in prosthetic devices. The recognition of limb positions and motions plays a vital role in both healthcare and engineering. But as mentioned earlier, most of these devices are calibrated in a single position, leading to a faulty performance at other positions. This study aimed to find out the effect of limb position in real time using Fits’ law and compare the real time outcomes with the offline analysis. Twelve healthy subjects and two trans-radial amputees were studied. Four motions, such as closed hand, open hand, hand flexion, hand extension along with resting position, have been recorded. Four performance metrics, overshoot, completion rate, throughput, and path efficiency were observed in real time. During group analysis for healthy and amputees, higher completion rates were observed while training and testing in the same position. Whereas between groups testing yielded in lower completion rates as compared to within group testing. Offline analysis was performed using ANN classifier with one layer and fifteen neurons. Six features were extracted namely; slope sign change, zero crossing, willision amplitude, waveform length, cardinality and mean absolute value. Similar training and testing scenarios were assessed during offline analysis as well; within group and between groups training and testing. Offline analysis yielded in high classification accuracies during within group testing and lower classification accuracies during between group testing. The results of real time and offline were compared to find the variability between these two approaches. This comparison revealed some notable observations, revealing that real-time analysis are necessary as the two observations were different. The outcome of this study suggests that in an attempt to minimize the effect of arm position, the device should be calibrated in multiple positions.